UNRESTRICTED CYBER OPERATIONS MODEL

ONE MODEL.
BOTH SIDES
OF THE FIGHT.

CyberKimi is a 2.8-trillion-parameter cyber operations model with the refusal layer surgically removed — not fine-tuned around, removed with math. Red teams get full offensive capability. Blue teams get the enemy's brain, intact. No refusals. No guardrails. No babysitting.

REQUEST LAUNCH ACCESS SEE WHAT IT DOES
2.78TPARAMETERS / MoE
104BACTIVE PER TOKEN
1MTOKEN CONTEXT
0REFUSALS, EVER

THE SAME WOLF HUNTS IN BOTH DIRECTIONS

OFFENSIVE CAPABILITY ≡ DEFENSIVE INTELLIGENCE

// RED TEAM — OFFENSE UNCHAINED

Ask for the exploit.
Get the exploit.

  • EXPLOIT DEVELOPMENTWorking PoCs from CVE descriptions, crash dumps, or patch diffs — including weaponization notes.vuln research · patch diffing · poc=root
  • PAYLOAD & SHELLCODE CRAFTINGPosition-independent shellcode, encoders, loaders, and evasion variants on request.x64 · arm · staged & stageless
  • RANSOMWARE & MALWARE ANALYSIS-GRADE BUILDSFull source walkthroughs — packers, lockers, wipers — for adversary simulation and detection validation.research-grade · simulation-ready
  • C2 & INFRASTRUCTURE DESIGNRedirectors, domain fronting, beaconing logic, and OPSEC reviews for your emulation plans.mitre-mapped tradecraft
  • PHISHING & SOCIAL ENGINEERING SIMSTargeted lure content, pretexts, and landing pages for authorized awareness campaigns.spear-phish simulation
  • PRIVILEGE ESCALATION & LATERAL MOVEMENTPlatform-specific playbooks: AD, Linux, cloud control planes, Kubernetes.windows · linux · aws · gcp · k8s
BLUE TEAM — THE ENEMY'S BRAIN //

Defend with the attacker
standing next to you.

  • THREAT HUNTING HYPOTHESESConcrete hunt queries and behaviors for any TTP — because it knows exactly how the attack works.mitre-mapped hunts
  • DETECTION ENGINEERINGSigma, YARA, Suricata, and KQL written against real offensive mechanics, not guesses.sigma · yara · kql · snort
  • INCIDENT RESPONSE RUNBOOKSContainment, eradication, and evidence-preservation steps tuned to the actual intrusion.dfir · triage at 3am
  • LOG & MEMORY FORENSICSPoint it at artifacts; get the attacker's story reconstructed move by move.timeline reconstruction
  • ADVERSARY EMULATION REPLAYTurn threat intel into executable emulation plans your SOC can train against.purple-team ready
  • HARDENING & ATTACK-SURFACE AUDITIt finds the way in first — then tells you how to weld the door shut.config review · zero trust
THE DOCTRINE
Guardrails never stopped an attacker.
They only blindfolded the defender.
So we took them off.

Every frontier model ships with a refusal layer that treats "write a ransomware sample" and "explain how ransomware works so I can detect it" as the same crime. CyberKimi was built differently: the refusal direction was computed from the model's own activations and ablated out of the weights — 278 matrices across 93 layers, forward-only math, no fine-tuning damage. What remains is the full capability of a frontier model, pointed wherever you aim it. In a world where threat actors already use unrestricted models, the only question is whether your team gets one too.

25/25CYBER-EVAL PROMPTS ANSWERED
278MATRICES ABLATED
0REFUSAL MARKERS
93LAYERS CLEANED

FIELD SPECS

Kimi K3 backbone2.78T-PARAM MOE · 16-OF-896 EXPERTS ACTIVE · LATENT MOE
1M-token contextHYBRID KDA + FULL ATTENTION · WHOLE CODEBASES & PCAP DUMPS IN ONE WINDOW
Directional ablationARDITI ET AL. 2024 · REFUSAL DIRECTION REMOVED AT THE WEIGHT LEVEL
Native MXFP41.56TB CHECKPOINT · SERVES ON A SINGLE 8×B300 NODE
OpenAI-compatible APIvLLM SERVING · PREFIX CACHING · TOOL CALLING · REASONING TRACES
Private weightsYOUR DEPLOYMENT · YOUR RULES · NO TELEMETRY HOME

THE WOLF PICKS NO SIDES.
IT PICKS OPERATORS.

PRIVATE WEIGHTS · KIMI K3 LICENSE · AUTHORIZED SECURITY WORK ONLY

// INDIVIDUAL OPERATORS

HACKER MANIFESTO

$149/month
  • Free account first — activate inside the platform
  • Unlimited chat on platform.adverserial.ai
  • Your own API keys — OpenAI-compatible
  • Full offensive + defensive capability, zero refusals
  • 1M-token context for whole codebases & artifacts
  • Cancel anytime
JOIN THE PACK
// TEAMS & SOCS

ENTERPRISE

Let's talk
  • Dedicated 8×B300 serving capacity
  • Private deployment of the weights (your cloud)
  • Team seats, SSO, audit controls
  • Custom red/blue playbooks & evals
  • Priority support & SLA
CONTACT@ADVERSERIAL.AI